I found the drawbacks of AI in research: 1. It can miss obvious information, which is very clear and falls under basic errors that I understand, until I remind it. 2. It cannot analyze from first principles. For example, if I propose a requirement, is this requirement necessary? Otherwise, it goes down the wrong path. But I am unclear about my requirements. This reminds me of Musk building rockets, where the first issue is questioning the problem and not answering stupid questions. 3. There are too many advertisements; the public network is filled with sales promotions. The chatter is a waste of time, corresponding to products that have never won online and never lost in reality, which AI may not be able to recommend. 4. The benefit is to quickly conduct basic research and then find experts. All my experiences and research involve finding the threshold for experts, roughly judging their level. Because I need to clarify where the core point is, and AI cannot handle that; it requires experts. To give an example from my own experience, I started systematically researching fitness in 2023, and over 90% of the content online is about bodybuilding training, which aligns with the commercial society of traffic. But a more effective way is powerlifting. My initial research was to understand that I needed to train for powerlifting but not bench press; I needed an online coach. So at the beginning, when communicating with the coach, I had various questions and adjustments I wanted to make. But actually, following the training plan, many issues disappeared directly. Moreover, the coach can easily tell you how to adjust, like lowering the barbell on your back a bit or standing a bit narrower. What is important and what is not really doesn't matter. AI is the kind that gives you a lot of analysis on trivial issues, and when you look at it, it seems somewhat reasonable, but you end up running further away on trivial matters. 🤣